Parameter Space Partitioning (PSP)
PSP Blob

Introduction

Parameter Space Partition (PSP) is a search method implemented in a computer algorithm that discovers qualtitative data predictions that a mathematical model could potentially generate. PSP does this by partitioning the model's parameter space. Each of partitioned regions corresponds to a unique data pattern the model can predict. PSP also provides an approximated volume of each of the discovered parameter regions.

PSP was developted as a tool for what we call Global Model Analysis as shown in Kim, Navarro, Pitt, and Myung (2004, available here) and Pitt, Kim, Navarro, and Myung (in press, available here). In these paper, the applicability of PSP was successfully demonstrated in comparing and analyzing two localist connectionist models of speech perception: TRACE (McClleland & Elman, 1986) and Merge (Norris, McQueen, & Cutler, 2000); and in analyzing a mathematical model of categorical learning: ALCOVE (Kruschke, 1992).

This website was designed to provide information for understanding and using the method of PSP.

List of Contents

  • How does PSP work?(revised from Appendices in Pitt et. al.)
  • Explains how the method of PSP works in finding a model's possible predictions by searching the model's parameter space. How a simple form of Markov chain can be useful in this context is discussed.

  • Java Illustration of PSP algorithm
  • Visually depicts how the PSP algorithm searches the parameter space to identify a model's predicted data patterns.

  • MATLAB codes and PSP manual
  • Commented MATLAB codes and the manual of PSP is provided.

    References

    Kim, W., Navarro, D.J., Pitt, M. A., Myung, I.J. (2004). An MCMC-based method of comparing connectionist models in cognitive science. Advances in Neural Information Processing Systems, 16, (937-944). MIT Press.

    Kruschke, J. K. (1992). ALCOVE: An exemplar-based connectionist model of category learning. Psychological Review, 99, 22-44.

    McClelland, J.L., & Elman, J.L. (1986). The TRACE model of speech perception. Cognitive Psychology, 18, 1-86.

    Norris, D., McQueen, J.M., & Cutler, A. (2000). Merging information in speech recognition: Feedback is never necessary. Behavioral & Brain Sciences 23, 299-370.

    Pitt, M. A., Kim, W., Navarro D. J., Myung I. J. (in press). Global Model Analysis by Parameter Space Partioning. Psychological Review.